Back to skill

Security audit

全方位智能股票分析v4

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed stock-analysis assistant, but it automatically reads a desktop knowledge folder and gives decisive trading guidance without enough user control or data-integrity safeguards.

Install only if you are comfortable with a Chinese-language stock-analysis skill that may automatically read files under ~/Desktop/股票知识库/ and produce buy/sell-style recommendations. Keep sensitive or unrelated files out of that folder, verify market data independently, and treat outputs as informational rather than investment advice.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:5
Finding

Implicit Access to a User-Controlled Local Knowledge Directory

Content
View full analysis
Key features: (1) conclusions first, (2) clear and decisive signals, (3) real-time intraday scanning, (4) automatically read ~/Desktop/股票知识库/. ``` The original directory path means a stock knowledge base on the user's desktop. ### Technical Analysis The Skill declares that it will automatically read an entire directory in the user's home directory. Access to local private files is not inherently necessary to obtain public market prices, fund-flow information, research reports, or news. The instruction does not establish: - Explicit user consent before each access. - File-level selection or an allowlist. - File-type and file-size restrictions. - A prohibition against including local content in web searches or generated reports. - Handling rules for credentials, personal data, or unrelated files found in the directory. This violates least-privilege principles at the agent level. The access remains constrained by the permissions of the process running the agent; no operating-system privilege escalation mechanism was identified. ### Attack Path 1. A user or another local application places sensitive, misleading, or attacker-controlled content in `~/Desktop/股票知识库/`. 2. The user submits a stock-analysis request, causing the Skill to activate. 3. The Skill automatically reads files from the directory without obtaining file-specific consent. 4. The local content enters the agent's working context. 5. That content may influence the investment analysis or be reproduced in the response. 6. If the agent incorporates the content into required web searches, portions could be disclosed to an external search provider. The reviewed instructions do not explicitly require exfiltration, so the final ...[truncated 735 chars]
Remediation
View remediation

other

Warning
Location
SKILL.md:194
Finding

Mandatory Disclaimer Can Falsely Claim Complete Data Verification

Content
View full analysis
All data in this analysis comes from real APIs > (Tushare/AkShare/yfinance), and every number has been verified. > Historical price changes are based on actual trading data. > This analysis is for reference only and does not constitute investment advice. > If an item cannot be verified, the credibility of that conclusion is reduced; > make your own judgment. ``` ### Technical Analysis The Skill simultaneously permits unavailable or unverified data and mandates a disclaimer asserting that every number has been verified. These states are logically incompatible. API failures, stale responses, symbol-resolution errors, partial historical coverage, or missing research data can trigger the documented unverified-data state. Nevertheless, the mandatory disclaimer directs the agent to make a blanket verification claim. The presence of a later warning about unverifiable items does not correct the false assertion that all numbers were verified. This is an output-integrity defect in a financial-analysis workflow. It can make incomplete or partially validated results appear more authoritative than the available evidence supports. ### Attack Path 1. A required market-data, research, fund-flow, or news source is unavailable or returns incomplete information. 2. The Skill proceeds under its documented missing-data rule and marks one or more items as unverifiable. 3. The mandatory disclaimer is appended to the response. 4. The disclaimer states that all data came from real APIs and every number was verified. 5. The user interprets the analysis as fully vali ...[truncated 908 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Vague Triggers

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger definition is extremely broad: any message containing a stock/ETF code or name plus an intent to get buy/sell advice may match many ordinary finance conversations. This can cause unintended invocation of a powerful skill that performs external data access and produces actionable financial recommendations, increasing the chance of inappropriate activation and user confusion.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation examples are broad enough to match ordinary stock-related conversation, which can cause the skill to trigger unexpectedly in general chats about securities. In a finance context, unintended activation is more sensitive because the skill presents explicit buy/sell-style analysis, increasing the chance of unsolicited financial guidance or use of external data/tooling when the user did not clearly request this skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

The skill name and description are entirely China-focused and written as a fixed Chinese experience, but the document does not state that language choice is optional or that the locale restriction is required for a region-specific compliance reason. Since the policy asks to flag forced language/locale behavior without user opt-in, this is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest sets the skill language to "zh-CN", which indicates a fixed Chinese locale. Under the policy, locale restrictions should either offer user choice or be clearly documented as justified; this file provides neither.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill name and description are entirely in Chinese and present the assistant as a stock-analysis tool for Chinese and international markets without indicating that language selection is optional. This can violate language/locale policy expectations because the skill appears to assume a fixed language rather than offering a user choice or documenting a justified locale restriction.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

All user-facing instructions, examples, and descriptions in the file are Chinese-only, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking or region-specific audience. This can violate language/locale policy when a skill implicitly forces a language without user opt-in.

Content

No source excerpt is available for this finding.